Medulloblastoma: Molecular Genetics and Animal Models

Abstract

Medulloblastoma is a primary brain tumor found in the cerebellum of children. The tumor occurs in association with two inherited cancer syndromes: Turcot syndrome and Gorlin syndrome. Insights into the molecular biology of the tumor have come from looking at alterations in the genes altered in these syndromes, PTC and APC, respectively. Murine models of medulloblastoma have been constructed based on these alterations. Additional murine models that, while mimicking the appearance of the human tumor, seem unrelated to the human tumor's molecular alterations have been made. In this review, the clinical picture, origin, molecular biology, and murine models of medulloblastoma are discussed. Although a great deal has been discovered about this tumor, the genetic alterations responsible for tumor development in a majority of patients have yet to be described.

Introduction

Medulloblastoma is a primary brain tumor that occurs in the cerebellum of children and young adults. The nomenclature for this tumor is somewhat controversial. The name “medulloblastoma” was given by Bailey and Cushing [1] in 1925; they suggested that these tumors arise from a hypothesized central nervous system (CNS) precursor cell called a medulloblast. Such a cell has never been identified, leading Rorke [2] to include medulloblastoma in a group of histologically similar CNS tumors, which she called primitive neuroectodermal tumors (PNETs). Some of the controversies regarding nomenclature have been addressed by recent studies of gene expression using array analysis (see below). Because these tumors arise in the cerebellum, patients present with symptoms of cerebellar dysfunction, including balance problems and incoordination. In addition, because the tumor frequently fills the fourth ventricle and blocks the normal circulation of cerebrospinal fluid (CSF), patients often present with symptoms and signs of hydrocephalus, including headache, vomiting, diplopia, and papilledema. Medulloblastomas are best demonstrated by magnetic resonance imaging (MRI), although characteristic findings can also be seen on computed tomography (Figure 1). Because these tumors have a propensity to spread through the CSF to distant CNS sites, patients should have MRI of the entire brain and spine. Medulloblastoma has an incidence of two to five cases per 10,000 population per year, resulting in about 240 new cases per year in the United States [3]. Most medulloblastomas arise sporadically, although medulloblastoma may arise rarely as part of an inherited cancer syndrome (see below).

Figure 1.

MRI of medulloblastoma. This T1-weighted, gadolinium diethylenetriamine pentaacetic acid (DTPA)-enhanced MR scan of a patient with a medulloblastoma demonstrates a mass (large area of white) in the cerebellum, which fills the fourth ventricle and displaces it anteriorly.

Grossly, medulloblastomas most often arise in the roof of the fourth ventricle. They grow to invade the cerebellar vermis and fill the ventricle, often invading through the ependyma in the floor of the ventricle to enter the brainstem (Figure 1). Less commonly, the tumor arises in the cerebellar hemisphere. Grossly, the tumors appear to be relatively well circumscribed. The tumor may contain areas of calcification or necrosis. Microscopically, the tumor cells are small with little cytoplasm (Figure 2A). They occur in sheets and may show signs of neuronal or glial differentiation. Histologic subtypes of medulloblastoma have been described and include the desmoplastic variant and the large cell variant. The desmoplastic variant is composed of islands of larger, pale cells in a sea of smaller, more typical medulloblastoma cells. In addition, an abundant collagenous matrix is present. In the large cell variety, the cells are larger and more pleomorphic. Microscopically, the tumor is invasive at its edges, although penetration into the surrounding cerebellum is somewhat limited. Immunhistochemical staining is frequently positive for synaptophysin and NeuN, indicating that neuronal differentiation and may be regionally positive for glial fibrillary acid protein (GFAP), indicating astrocytic differentiation in some parts of the tumor.

Figure 2.

(A) Photomicrograph of a human desmoplastic medulloblastoma, demonstrated as small, tightly packed cells with little cytoplasm. A pale island is seen in the center of the image as an area of less tightly packed cells. Hematoxylin and eosin staining, x400. (B) Photomicrograph of murine tumor of the cerebellum from a Ptc^+/- mouse. Note the similarity to the human tumor. Hematoxylin and eosin staining, x400. (Image courtesy of Dr. Cynthia Wetmore, Rochester, MN.)

Current Therapy of Medulloblastoma

Currently, patients with medulloblastoma are best treated with surgical removal of the tumor, and adjuvant radiation therapy and chemotherapy (for review, see Refs. [4–7]). Many prospective, randomized trials have demonstrated that patients do better if most—if not all—of the tumor is removed. Radiation therapy, at a dose of at least 55 Gy to the tumor, has also been shown to prolong survival and result in cures. Because the tumor has a propensity to spread in the CSF, an additional 35 Gy is given to the entire brain and spinal cord. The devastating effects on intellect caused by the radiation therapy in young children have been well documented. The younger is the child at the time of irradiation, the worse is the intellectual outcome. For this reason, children under the age of 3 years are often treated with chemotherapy alone. Agents with demonstrated efficacy against medulloblastoma include DNA alkylating agents such as ethylnitrosoureas (BCNU and CCNU) and the platinum derivatives (cis-platinum and carbo-platinum). Chemotherapy is also used in conjunction with radiation therapy in older patients, especially those with CSF metastases at presentation.

Cerebellar Development

The cerebellum develops from neuronal precursors located in the rhombic lip of the fetal brain (for review, see Ref. [8]). These precursors migrate tangentially to form the external granular cell layer (EGL) of the developing cerebellum. Purkinje neurons and Bergman glia arise from precursors in the subventricular zone and migrate toward the EGL. In the EGL, the granular cell precursors continue to proliferate in the outer zone. Postmitotic neurons move to the inner zone of the layer and then migrate along Bergmann glial fibers to finally reside in the internal granular cell layer (IGL). The EGL eventually disappears as all cell division ceases and all postmitotic neurons move to the IGL. The number of neurons that finally reside in the IGL is large. In fact, there are more neurons in the IGL than in the rest of the brain combined.

The sonic hedgehog (SHH) signaling pathway plays an essential role in cerebellar development [9,10]. The receptor for SHH, patched (PTC), is a membrane-associated protein containing 12 transmembrane domains [11,12]. Associated with PTC in the membrane is the effector molecule, smoothened (smo) [13]. PTC appears to function to inhibit signaling by smo. Binding of SHH to PTC releases this inhibition, resulting in activation of the intracellular components of the pathway. The main effector of SHH in the cell may be gli1, a transcriptional activator [14] (Figure 3). In the developing cerebellum, PTC is expressed by neuronal precursors in the EGL [9,10]. SHH is produced by Purkinje neurons that lie beneath the EGL. In vitro, SHH has been shown to be a potent mitogen for EGL precursors. Blocking of SHH signaling in vivo leads to hypoplastic cerebella in which granule neurons are greatly reduced or absent. These results strongly suggest that EGL neuronal precursors are stimulated to divide by SHH signaling. The factors that lead to differentiation and migration of the postmitotic neurons are not clear at this time, although β-FGF has been shown to block the SHH-induced proliferation of EGL cells in vitro. Figure 4 summarizes role of the SHH pathway in the development of the cerebellum.

Figure 3.

The SHH/PTC pathway. By binding to the membrane-bound PTC receptor, SHH removes the inhibition of v-smo mediated by PTC. This ultimately results in increased gli-mediated transcription. Su-fu and PKA are downstream inhibitors of this process.

Figure 4.

The role of SHH in cerebellar development. Granule neuronal precursors (A–D) migrate tangentially from the rhombic lip and may use the SHH pathway in transient autocrine manner. Purkinje neurons and later-born Bergmann glia (B) derive from the ventricular zone and migrate toward the EGL. SHH from the Purkinje neurons induces Bergmann glia maturation (C). In the later EGL, granule neuronal precursors proliferate in the outer zone, utilizing SHH secreted from Purkinje neurons. At the same time, mature glia send their extensions toward the inner EGL (D) and these or other cortical cells may provide factors that promote the differentiation if granule neurons, antagonizing the effects of SHH. Granule cells then migrate on glial fibers across the molecular and Purkinje layers to form the IGL. Maintained autocrine SHH signaling in the EGL (E) may result in the development of cerebellar tumors. (Reprinted with permission from Ref. [3].)

Other signaling pathways have been implicated in the development of the cerebellum. Both neurotrophins and their receptors have been found to be expressed in the developing cerebellum. The temporal pattern of expression suggests that brain-derived neurotrophic factor (BDNF) and its receptor, trkB, may be involved in the division of cells in the EGL, and that NT3 and its receptor, trkC, may be involved in the terminal differentiation of these cells into internal granular layer neurons [15,16]. This model is of interest in light of data suggesting that the expression of trkC in medulloblastoma is a marker for better prognosis [17,18].

Medulloblastoma: Cell of Origin

Medulloblastomas are observed to differentiate along glial and neuronal pathways in situ, suggesting that these tumors are derived from primitive, pluripotent, neuroepithelial stem cells. This conclusion is supported by studies of PNET cell lines that demonstrate expression of specific, developmentally regulated proteins in PNETs [19]. Medulloblastomas have been demonstrated to express zic, a gene normally expressed only in the EGL of the developing cerebellum and its derivatives, suggesting that medulloblastoma arises from EGL precursor cells [20,21]. Given the rapid proliferative capacity of EGL precursors and the pattern of gene expression seen, EGL cells and their derivatives seem, by far, the most likely origin of medulloblastoma. Other cerebellar cells, such as Purkinje cells, basket neurons, or glial cells, are unlikely to be cells of origin for medulloblastoma. Evidence that other stem cells in the cerebellum may give rise to medulloblastoma comes from murine models of human medulloblastoma (see below). For example, a model of medulloblastoma has been generated using the GFAP promoter to drive expression of the RecA recombinase, resulting in tissue-specific inactivation of RB1 [22]. In this work, cells expressing GFAP were identified in the developing cerebellum, although the vast majority of EGL precursors did not express GFAP.

Medulloblastoma: Karyotypic Abnormalities

Only one karyotypic abnormality has been found to be typical of medulloblastoma—the presence of an isochromosome 17q, present in about 50% of tumors [23]. The breakpoint has been localized to 17p11.2, but no tumor-specific gene rearrangement has been identified. Despite multiple studies, no tumor-suppressor gene that can be implicated in the development of medulloblastoma has been found on chromosome 17p. Specifically, no alteration in p53 has been identified with more than 100 tumors investigated to date. The breakpoint for the rearrangement has been mapped to 17p11.2 [24]. Other less common karyotypic abnormalities, including loss of heterozygosity (LOH) on chromosome 9q, have been identified in about 20% of medulloblastomas. Interestingly, the loss of 9q in medulloblastoma has been correlated with the desmoplastic subtype [25].

Analyses of Gene Expression in Medulloblastoma

Two recent reports of gene expression in medulloblastoma have added insights to the basic biology of these tumors. In the first of these, medulloblastoma samples were divided into two groups, depending upon whether CSF dissemination of tumor was seen at the time the patient presented with the tumor [26]. Gene expression was analyzed on Affymetrix G110 cancer arrays, which are enriched for genes thought to be important in cancer biology. Fifty-nine genes that had increased expression in the metastatic tumors compared to the nonmetastatic tumors were identified. An additional 29 genes were identified, which had decreased expression in the metastatic tumors. Two of theses genes, PDGFRα and SPARC, were shown by immuohistochemical staining to be expressed differentially between metastatic and nonmetastatic tumors. In addition, antibodies to PDGFRα blocked tumor cell migration in in vitro assays. These investigators were able to use the pattern of gene expression to predict which tumors presented with CSF metastases in a blinded group of tumors. These results suggest that there may be genes important in the progression of medulloblastoma, or that alterations in different pathways may be responsible for the differences in prognosis seen between tumors presenting without and with CSF dissemination.

In the second study, a group of “embryonal tumors” of the CNS, including medulloblastomas, extracerebellar PNETs, atypical teratoid/rhabdoid tumors, and malignant gliomas, was presented [27]. Gene expression was analyzed on Affymetrix HuGene FL arrays, containing 5920 known genes and 897 expressed sequence tags. The first finding from this data set was that each tumor type could be easily distinguished from the others based on gene expression pattern. Especially interesting was the ability to distinguish medulloblastomas from other CNS PNETs, suggesting that the lumping of these tumors together, as proposed by Rorke, may be inappropriate. The medulloblastomas often expressed genes normally expressed in the EGL, lending credence to the hypothesis that EGL precursors represent the cell of origin of medulloblastoma. Second, these investigators were able to distinguish “classic” medulloblastomas from desmoplastic medulloblastomas. The important genes with increased expression in the desmoplastic tumors were those involved in the SHH/PTC pathway (see below). Lastly, using an eight-gene model, these investigators were able to accurately predict outcome in their patients with medulloblastoma. Genes associated with favorable clinical outcome included TRKC and other genes characteristic of cerebellar differentiation. In contrast, these genes were underexpressed in tumors with poor prognosis and genes involved in cell proliferation were overexpressed.

Congenital Cancer Syndromes and Medulloblastoma

Medulloblastomas may occur in association with two different inherited cancer syndromes: Gorlin syndrome and Turcot syndrome. Nevoid basal cell carcinoma syndrome (NBCCS), also called Gorlin syndrome or basal cell nevus syndrome, is an autosomal dominant disorder [28]. Affected individuals develop multiple basal cell carcinomas, multiple odontogenic keratocysts of the jaws, palmar and plantar dyskeratoses, and skeletal anomalies, especially rib malformations. In addition, at least 40 cases of medulloblastoma have been reported in patients with this syndrome, indicating that about 3% of Gorlin patients develop this tumor [29,30]. The gene for Gorlin syndrome has been mapped to chromosome 9q22.3 [31,32]. Two studies have reported loss of genetic markers mapped to 9q in medulloblastoma. In the first study, 16 patients were examined with 12 microsatellite markers mapping between 9q13 and 9q34 [33]. Two tumors (12.5%) showed LOH with microsatellite markers in this region. In the second study, medulloblastomas from 20 patients, 17 with sporadic tumors and 3 with NBCCS, were investigated with seven microsatellite markers mapped to 9q22.3 to 9q31 [34]. Both informative tumors from patients with NBCCS showed LOH for markers in the region; the third patient was not informative. Three of the 17 sporadic tumors also showed LOH on 9q. Interestingly, all three of the tumors from patients with NBCCS in this study were designated desmoplastic medulloblastomas. The other three tumors with LOH on 9q were among six desmoplastic tumors in the sporadic group. Thus, all of the tumors with LOH on 9q in this study were desmoplastic, raising the possibility that an NBCCS gene mutation is involved in the development of this subclass of tumor.

The gene at 9q22.3 responsible for NBCCS has been identified as the PTCH gene, the human homolog of the Drosophila patched gene [35,36]. The Drosophila gene encodes a protein with 12 putative transmembrane domains; it may function as a receptor or transporter [37,38]. The protein has an essential role in correct patterning of larval segments and imaginal discs during adult fruit fly development; a similar role in humans may explain the congenital anomalies associated with NBCCS.

PTCH mutations have been described in patients with NBCCS and in spontaneous basal cell carcinomas. A large number of such mutations have been reported, including single base substitutions, insertions ranging from a single base to 300 bases, and deletions ranging from a single base to 37 bases [35,36,39–42]. The described mutations are fairly evenly distributed throughout the PTCH gene; no “mutational hot spots” have been identified.

Turcot syndrome is a hereditary disorder in which affected individuals have multiple colonic polyps and a brain tumor, either glioblastoma multiforme or medulloblastoma [43]. In one study, mutations in the APC gene were identified in the group of patients with Turcot syndrome who developed medulloblastoma [44]. The relative risk for developing a medulloblastoma in patients with Turcot syndrome and an APC gene mutation is 92 times that in the general population.

APC functions as a key regulator in a complex developmental pathway (Figure 5). In the cytoplasm, APC associates with at least seven proteins, including β-catenin, glycogen synthase kinase 3β (GSK-3β), axin1 and axin2, β-TrCP, the B6 subunit of the PP2A phosphatase, and hDLG [45–55]. The control of the levels of free β-catenin in the cytoplasm by APC defines the role of APC as a tumor suppressor. Normally, free levels of β-catenin are low, as binding of β-catenin by APC sequesters β-catenin and targets the protein for degradation. APC only binds β-catenin when β-catenin is hyperphosphorylated. β-Catenin is phosphorylated by GSK-3β, a serine/threonine kinase. β-Catenin contains four phosphorylation sites for GSK-3β, three serines and one threonine, all encoded in exon 3 of the β-catenin gene. When APC is inactivated by mutation (e.g., in colon carcinoma), levels of cytoplasmic β-catenin rise. Free β-catenin associates with members of the Tcf family [56–58]. Four family members have been described; all are transcription-regulating proteins with DNA-binding activity. When β-catenin associates with Tcf, the complex moves to the nucleus and upregulates the expression of genes that increase the rate of cell division, either by stimulating cell proliferation or by inhibiting apoptosis.

Figure 5.

The wnt signaling pathway. Binding of wnt to its receptor, frizzled, leads ultimately to translation of β-catenin and Tcf into the nucleus, resulting in transcription of genes controlled by this transactivator. The large complex involved in regulating free cytoplasmic concentrations of β-catenin contains axin1 and axin2, APC, and GSK-3β. Phosphorylation of β-catenin by GSK-3β in the complex leads to degradation of β-catenin by the ubiquitin system.

Genetic Alterations in Medulloblastoma

Based on the germline gene alterations found in the inherited syndromes described above, sporadic medulloblastomas have been investigated for alterations of the involved genes. Inactivation of the PTCH locus by deletion and mutation has been found in about 10% of sporadic medulloblastomas, suggesting that PTCH functions as a classic tumor suppressor in this subset of tumors [59–64]. An analysis of the SHH/PTCH pathway reveals that other genes in the pathway might be altered to give a phenotype similar to that caused by PTC inactivation. An investigation looking for these alterations in 15 other pathway genes has revealed only very rare mutations in v-smo and suppressor of fused [su(fu)] [60,65,66]. This finding suggests that there is something unique about PTCH inactivation, or that the PTCH locus is relatively easy to inactivate through mutation or deletion.

Similarly, the APC gene has been investigated for inactivation in sporadic medulloblastoma. Surprisingly, in light of the association between germline APC gene mutation and medulloblastoma seen in Turcot syndrome, APC gene mutations have rarely been identified in spontaneously occurring medulloblastomas. In one study, 47 sporadic medulloblastomas were examined for mutations in the APC mutation cluster region, comprising 10% of the gene, where more than two thirds of the mutations seen in colorectal carcinoma occur [67]. No tumor had an APC mutation in this region. However, the entire gene was not investigated in this report. Interestingly, in the two medulloblastomas in this study that were removed from patients with Turcot syndrome, the germline mutation was identified in tumor DNA, but no mutation was seen in the other APC allele. In a second study, DNA from 23 medulloblastomas was examined for deletions in the region of the APC gene [68]. No LOH was found in this region in any tumor. However, the four microsatellite markers used in this study, although the closest to APC available at the time, were at least 30 to 70 kilobases from the APC locus. In a third study, two tumors among 46 medulloblastomas were found to have APC mutations [69]. However, the mutations identified were very conservative (i.e., alanine to valine and valine to isoleucine). The mutations did not occur in defined functional domains of the protein. The functional consequences of these mutations are not clear. Taken together, these results indicate that APC mutations in sporadic medulloblastoma are quite rare and may have no functional consequence.

Because β-catenin/Tcf complexes are translocated to the nucleus, nuclear localization of β-catenin, demonstrated immunohistochemically, has been used as a marker for tumors with active β-catenin/Tcf transcription [70–73]. Nuclear localization of β-catenin occurs in tumors with inactivated APC or with oncogenic β-catenin mutations, suggesting that this finding identifies tumors with increased β-catenin/Tcf transcription, regardless of the mechanism responsible. Immunohistochemical staining for β-catenin has been performed in medulloblastoma [74]. Nine of 51 sporadic medulloblastomas were found to have nuclear localization of β-catenin. This result suggests that at least 20% of medulloblastomas have alteration in the control of β-catenin levels other than APC inactivation.

In colon carcinoma, free β-catenin levels may be increased by two mechanisms. The first, inactivation of APC, has been described above. The second mechanism involves mutation of the β-catenin gene (locus CTNNB1) itself [75]. Mutations of β-catenin that alter the GSK-3β phosphorylation sites in exon 3 have been described. These missense mutations change the serines or threonines that are GSK-3β phosphorylation sites to cysteine, and prevent β-catenin from being completely phosphorylated. Cotransfection experiments using a reporter construct and mutated β-catenin demonstrated that these mutations exerted a dominant effect, rendering β-catenin/Tcf transcription resistant to APC-mediated downregulation. Transcriptional activities of β-catenin/Tcf reporter constructs in cells with wild-type APC transfected with mutant β-catenin were at least six times higher than in the same cells transfected with wild-type β-catenin. Gel shift analysis also demonstrated that free β-catenin was constitutively bound to Tcf-4 in nuclear extracts from cells containing mutant β-catenin, even in the presence of wild-type APC. Functionally similar mutations in which exon 3 has been deleted from the β-catenin gene have been described [76]. Taken together, these data indicate that mutations eliminating a GSK-3β phosphorylation site from β-catenin result in a protein that is no longer regulated by APC, but that continues to function as a transactivator when bound to Tcf. This model predicts that mutated β-catenin functions as a dominant oncogene, and this has been shown to be the case. As predicted by this model, inactivating mutations of APC and oncogenic mutations in β-catenin occur in nonoverlapping subsets of colon carcinomas. The ubiquitin-binding region of β-catenin is also in exon 3. Mutation of this region, leading to lack of β-catenin ubiquitization, has also been shown to be oncogenic.

Given that germline APC mutations in Turcot syndrome give rise to an increased incidence of medulloblastoma and given the rarity of APC mutations in sporadic medulloblastoma, sporadic medulloblastomas have been screened for oncogenic mutations in β-catenin, as these mutations result in the same phenotype as APC inactivation. Exon 3 of β-catenin, the location of all GSK-3β phosphorylation sites, was investigated by direct sequencing in 67 sporadic medulloblastomas [77]. Five tumors were found to have β-catenin mutations. In three cases, the alteration resulted in the substitution of a cysteine for a serine at a GSK-3β phosphorylation site. Sequencing of exon 3 in the constitutional DNA from these patients (available for two of the three patients) failed to show a mutation, indicating that the mutation identified is tumor-specific. Two additional mutations were found, both of which altered an amino acid in the ubiquitin-binding region, also in exon 3. Thus, 5 of 67 (7.5%) tumors had oncogenic mutations in β-catenin. As expected, those tumors with β-catenin mutations also had nuclear localization of β-catenin. Similar results have been reported recently from another group [69], who reported that of 46 medulloblastomas, four (8.7%) were found to have oncogenic mutations in β-catenin. The finding that one quarter of medulloblastomas have nuclear localization of β-catenin, whereas only 7.5% contain oncogenic β-catenin mutations, suggests that alterations in this pathway—other than β-catenin mutation that result in increased β-catenin/Tcf transcription—may be occurring in a subset of medulloblastoma.

As mentioned above, binding of β-catenin by APC leads to degradation of β-catenin. At least two other proteins are involved. In humans, these proteins have been called axin1 and axin2 (Figure 5). The protein products of both of these genes associate with β-catenin, APC, and GSK-3β. Both proteins appear to facilitate phosphorylation of β-catenin by GSK-3β. In fact, in the absence of axin, GSK-3β phosphorylates β-catenin poorly. A role for axin1 in bringing GSK-3β and β-catenin together by mutual binding has been proposed. Inactivation of axin1 or axin2 would be expected to increase levels of free β-catenin through decreased degradation and, thus, increase β-catenin/Tcf-mediated transcription. Indeed, axin1 has been shown to be mutated in a subset of hepatocellular carcinomas [78]. Hepatocellular carcinoma cell lines with mutated axin1 demonstrated nuclear localization of β-catenin. Replacement of axin1 activity by transfection with wild-type axin1 resulted in apoptosis in these cell lines.

Medulloblastomas have been examined for axin1 mutations [79,80]. In the first study, of 86 tumors, eight were found to have mutations in axin1. There were seven deletions that went from intron to intron and included at least two exons and one point mutation. In four instances, an LOH analysis revealed no evidence of loss of the wild-type allele. However, the second study revealed these large deletions to be a PCR artifact. The actual frequency of axin1 mutations is low, with missense mutations found in 2 of 39 tumors examined. These results suggest that axin1 mutations are rare in medulloblastoma.

A third genetic alteration found in medulloblastoma is c-myc or N-myc amplification. Amplification of myc genes is rare, occurring in only 4% of medulloblastomas. Interestingly, c-myc amplification appears to correlate with the large cell phenotype, which carries a particularly grim prognosis.

Other Growth Factor Pathways in Medulloblastoma

Gilbertson et al. [81–83] have investigated the expression of erbB family members in medulloblastoma. They have shown that erbB2 and erbB4 are frequently expressed together in medulloblastoma. Interestingly, erbB4, but not erbB2, was expressed in the developing cerebellum. Expression of erbB2 and erbB4 was associated with simultaneous expression of neuregulin1-α, suggesting the possibility of an autocrine loop in tumors with expression of all three proteins. Indeed, erbB2/erbB4 dimerization was identified in tumors. This group has also shown that novel splice variants of erbB4 are found frequently in medulloblastoma. Significantly, coexpression of erbB2 and erbB4 indicated a worse prognosis and that expression of these receptors with neuregulin1-α was associated with CSF dissemination at presentation.

Activation of IGF-1R has also been demonstrated in medulloblastoma cell lines [84]. Autophosphorylation of this receptor and induction of c-fos expression in the presence of exogenous IGF-1 have been shown, indicating a functional receptor. Tumor growth was inhibited by an anti-IGF-1R antibody that interferes with ligand binding. Similar results have been reported by others [85].

Murine Models of Medulloblastoma

In the past, xenografts of established medulloblastoma cell lines in nude mice have been used as the standard model for this tumor in translational research. Unfortunately, these models suffer from a number of problems. First, the two most commonly used cell lines are not typical of the human tumor. One of these, TE671 was used for a number of years until proven to be a sarcoma rather than a medulloblastoma [86]. The second, Daoy, has genetic alterations (such as absence of wild-type p53 and homozygous deletion of the CDKN2 gene, which encodes the tumor suppressors p16 and p15) that are more characteristic of other types of brain tumors than of medulloblastoma [87,88]. This raises the issue of the use of cell lines to model medulloblastoma. PNETs have proven difficult to establish in culture. Only a few cell lines have been well characterized. Thus, it is possible that the established cell lines are not representative of PNET as a whole. Established cell lines could represent a subpopulation of PNET with specific mutations that allow for growth in vitro, or in vitro passage could select for mutations that occur during initial passages that allow the tumor to grow in vitro.

Because of the limitations of xenograft models, and in an attempt to replicate the human tumor, two laboratories have developed ptch knockout mice [89,90]. Homozygous inactivation of ptch causes embryonic lethality due to CNS system defects including failure of neural tube closure; overgrowth of head folds, hindbrain, and spinal cord; and cardiac defects [89]. Hemizygous ptch mice (ptch +/-) have many of the features of Gorlin syndrome, including skeletal abnormalities, neural tube closure defects, a generalized overgrowth, and a predisposition to tumor development. Indeed, about 30% of these mice develop tumors in the cerebellum that resemble human medulloblastoma histologically (Figure 2B). As expected, expression of gli1 is increased in these tumors, compared to expression in normal cerebellum, suggesting activation of the SHH signaling pathway in these tumors. Surprisingly, two independent studies have failed to demonstrate any alteration in the wild-type allele in the murine tumors [91,92]. Given the incidence of the tumors and the retention of the wild-type allele, the possibility of epigenetic silencing of the wild-type allele or mutation in other genes must be considered in this model. Epigenetic silencing is unlikely, as both studies demonstrate expression of wild-type mRNA in the tumors. However, a recent study has suggested that the wild-type ptch allele is silenced through methylation [93]. Unlike the two studies mentioned above, these investigators were unable to identify wild-type ptchmRNAin tumors from ptch +/- mice. They show that treatment of tumor cell lines with an inhibitor of methylation results in downregulation of the SHH/ptch pathway in tumor cell lines. How to reconcile this result with the identification of wild-type message in tumors cells by in situ hybridization is problematic.

The issue of other genes cooperating with hemizygous inactivation of ptch has been investigated by Wetmore et al. [94]. When the hemizygous ptch alteration is placed in a p53-null background, more than 95% of the mice develop medulloblastoma. In addition, the tumor arises earlier on a p53-null background when compared to the timing of tumor development when normal p53 is present. The increase in tumor development was seen only on a p53-null background and was not associated with an APC hemizygous background (Min +/-) or a p19/ARF hemizygous inactivation background. Given that p53 alterations are rarely seen in sporadic human medulloblastoma, these results suggest the possibility that the increased genomic instability of the p53-null background may lead to mutations in genes that cooperate with the hemizygous ptch alteration in the formation of these tumors.

Another murine model of medulloblastoma that depends upon the overactivity of the SHH pathway has been developed by Weiner et al. [95]. These investigators have injected the early cerebellar precursors of E13.5 mice in utero with a SHH-containing retroviral vector, using a sophisticated ultra-sound guided technique. When examined at P14 or P21, the injected mice have retained the EGL, suggesting that continued expression of SHH in the EGL results in persistent mitosis or prevents differentiation. In addition, the cerebellum of injected mice contains nodules of mitotically active tissues that resemble medulloblastoma. When examined at 13 weeks of age, injected mice have larger tumors that are exerting mass effect on the cerebellum. Interestingly, given the role of gli1 in the SHH pathway, identical findings were seen when fetuses of gli1-null mice were injected with the retrovirus. This result suggests that gli1 is not required for tumor development in this model. Perhaps another gli family member is sufficient.

A similar murine model of medulloblastoma has been developed using the replication-competent ALV splice (RCAS)/tv-a system [96]. In this model, a mouse line expressing the avian retroviral vector receptor, tv-a, under the control of the nestin promoter was used. Neonatal mice were injected with an avian RCAS acceptor vector into the cerebellum. Three of 32 mice (9%) developed medulloblastoma. An additional 5 of 32 developed multifocal persistence and hyperproliferation of the EGL. When the nestin/tv-a mice were injected with both a SHH/RCAS vector and a c-myc/RCAS vector, 9 of 39 mice (23%) developedmedulloblastoma. These results suggest that a nestin-expressing cell in the neonatal cerebellum is able to serve as a precursor to medulloblastoma. Whether the cooperation between SHH and c-myc relates to the human tumor remains to be investigated.

An interesting model of medulloblastoma has been reported in lig4-null mice. The lig4 gene encodes a protein that participates in the repair of DNA double-strand breaks by the nonhomologous end joining (NHEJ) complex. Mice null for lig4 die in late embryonic development because of extensive apoptosis of neurons in the CNS [97]. However, p53-null, lig4-null mice survive [98]. The double-null mice develop both lymphomas and medulloblastomas [99]. These investigators attribute the DNA strand breaks to “genotoxic stress” and suggest that the development of medulloblastoma in the double-null mice is related to mutations in undefined genes caused by improper repair of DNA damage. Which genes these may be has not been determined.

A second model of medulloblastoma in mice that combines the p53-null state with defective DNA repair has also been described [100]. In this model, the DNA repair gene inactivated is poly(ADP-ribose polymerase) (PARP-1), a DNA strand break-sensing gene that is activated as an early response to DNA damage. When p53-null, PARP-1-null mice were investigated, they were found to have a high frequency of a cerebellar tumor resembling human medulloblastoma. The progression of these tumors was associated with reactivation of expression of Math-1, a transcription factor expressed in the EGL. Like the preceding model, this model suggests that disruption of DNA repair in the EGL can give rise to medulloblastoma. The genes mutated in this model have also not been identified.

Two other murine models of medulloblastoma that involve the inactivation of p53 and pRB1 have been developed. In an attempt to generate astrocytic tumors, Marino et al. [22] constructed a mouse with conditional knockout of the RB1 gene using the Cre/LoxP system. The Cre recombinase was placed under control of the GFAP promoter to establish Cre expression in astrocytes. When the GFAP-Cre mice were crossed with Rb^Loxp/LoxP mice, inactivation of Rb in cells that express GFAP occurs. When these conditions of RB1 knockout mice were combined with a p53-null background, the resulting mice developed lymphoma, sarcomas, and tumors of the cerebellum resembling medulloblastoma. These investigators were able to demonstrate GFAP expression in the EGL precursors in the cerebellum of normal mice and inactivation of RB1 in the same cells in the GFAP-Cre, Rb^Loxp/LoxP mice. These results suggest that the tumors are indeed arising from EGL precursors. In addition, alteration of cell cycle control by RB1 inactivation is implicated in the development of this tumor in mice.qqq

A different model of medulloblastoma that also involves inactivation of pRB and p53 has been developed by Krynska et al. [101]. These investigators constructed a transgenic mouse line with the early sequences from JC virus that contain the early promoter and the large T antigen gene. These mice developed tumors of the cerebellum that resemble medulloblastoma. This model shares functional features with the model of Marino et al. in that large T antigen binds and functionally inactivates both p53 and pRb. In addition, tumors expressing T antigen contained evidence of increased activity of the β-catenin/Tcf transactivator, which has been implicated in the development of a subset of human tumors.

Given that human medulloblastomas rarely contain either p53 or pRB alterations, results from these two models may not be easily extrapolated to the human condition. Possibly, these models indicate that EGL precursors are easily transformed by a number of possible alterations, only some of which occur in humans. Alternatively, the key event in the generation of these tumors may not be the alteration in control of the cell cycle, but may rather relate to the activation of events such as β-catenin/Tcf-regulated transcription that have been seen in the human condition.

In summary, a number of murine models of medulloblastoma have been described. Some of these are the result of genetic alterations that are similar to those defined in human tumors, whereas others seem unrelated to the human condition. Hopefully, work with these models will lead to new insights into human medulloblastoma and to better therapies directed by the molecular alterations. Additional models with alterations in β-catenin or other wnt pathway members have yet to be described.

Summary

A great deal of information about the genetic alterations leading to the pediatric brain tumor medulloblastoma has been reported. Subsets of tumors with alterations in genes important in CNS development have been defined and murine models based on some of these have been generated. Nonetheless, the percentage of tumors that have identified gene alterations is less than 50% of the total. Clearly, more work is needed to define what alterations are responsible for uncontrolled cell proliferation in a majority of these tumors.

Glossary

• Cerebellum

  A part of the brain involved in control of coordination and balance. Made up of two laterally placed hemispheres connected by a central portion called the vermis

• CSF

  A clear, colorless fluid made by the brain. Fluid is made in chambers inside the brain (called ventricles), flows through channels in the brain to get outside the brain, bathes the surface of the brain and spinal cord, and is absorbed into the venous bloodstream

• Hydrocephalus

  When circulation of CSF is blocked, synthesis of the fluid continues, leading to accumulation of the fluid in the chambers in the brain. The chambers enlarge and exert pressure on the surrounding brain

• Pleomorphic

  Having a varied appearance related to cell size, cell shape, and nuclear size

• Synaptophysin

  A protein component of the vesicles that contains neurotransmitters in neurons

• GFAP

  An intermediate filament protein whose gene expression is restricted to astrocytes, one of the three main cell types in the brain

• SHH

  An extracellular signaling molecule. The protein is cleaved into two parts; the active amino-terminal portion is also cholesterolated. Its homolog in Drosophila is hedgehog

• PTCH

  The membrane-bound receptor for SHH. The protein contains 12 transmembrane domains and associates in the membrane with v-smo. The Drosophila homolog is patched

• v-smo

  The effector of SHH signaling. PTCH exerts inhibition of v-smo activity; this inhibition is released by SHH binding to PTCH. The Drosophila homolog is smoothened

• gli1

  An intracellular effector of SHH signaling. There are three members of the gli family. The Drosophila homolog is cubitus interruptus

• Neurotrophins

  A family of growth and differentiation factors related to nerve growth factor (NGF) and including NGF, BDNF, neurotrophin 3 (NT3), and neurotrophin 4 (NT4)

• Neurotrophin receptors

  The trk family of tyrosine kinases. TrkA has the highest affinity for NGF, trkB for BDNF, and trkC for NT3

• PDGFRA

  One of two tyrosine kinase receptors for platelet-derived growth factor (PDGF)

• APC

  The adenomatous polypi coli gene. The protein product of this gene is involved in the degradation of phosphorylated β-catenin

• β-Catenin

  A transcriptional transactivator involved in wnt signaling

• Tcf

  The T-cell factor family contains four members. They provide the DNA-binding specificity for the β-catenin/Tcf transactivator

• axin1

  A gene whose protein product is part of the APC complex that degrades phosphorylated β-catenin

• EGFR

  The membrane-associated tyrosine kinase receptor for epidermal growth factor and transforming growth factor-α

• IGF-1R

  The membrane-associated tyrosine kinase receptor for insulin-like growth factor 1

• Xenograft

  Growth of tissue from one species in a different species